conversational mental verbs in english song lyrics: a

Conversational Mental Verbs in English Song Lyrics:

A Corpus-Driven Analysis

Flora Goyak1, Mazura Mastura Muhammad2*, Farah Natchiar Mohd Khaja3, Muhamad Fadzllah

Zaini4, Ghada Mohammad5

1234Faculty of Language and Communication, Universiti Pendidikan Sultan Idris, 35900 Tanjong Malim, Perak, Malaysia

[email protected] [email protected]

[email protected] [email protected]

5The English Language Centre, University of Bahrain, Bahrain. [email protected] *Corresponding Author

https://doi.org/10.24191/ajue.v17i1.12619

Received: 25 November 2020 Accepted: 25 January 2021

Date Published Online:8 March 2021 Published: 8 March 2021

Abstract: This corpus-driven study explores the linguistics phenomenon of mental verbs in English song lyrics from 1960s until 2000s. This study aims to identify the frequency distributions of lexical verbs, mental verbs, and to analyze the language uses of mental verbs in the Diachronic Corpus of English Song Lyrics (DCOESL). Quantitative and qualitative methods of analysis were applied. First, quantitative data covering frequency distributions of general verbs was produced via LancsBox. Top three mental verbs in song lyrics were selected for analysis and discussion. The frequency distributions of mental verbs and collocations were produced via LancsBox. Collocational patterns were illustrated through collocational graphs constructed via LancsBox. Frequency distributions of mental verbs were compared to reference corpus Contemporary Corpus of American English (COCA) for the purpose of generalizing the findings from this study as representative of English language. The statistical data were submitted for four statistical tests of significance namely Chi-square, Mutual Information, Log-likelihood, and t-score. Second, qualitative data was composed of corpus annotations. Corpus annotations were conducted via CLAWS for assigning part-of-speech C7 tagset to identify verbs. Semantic categories of mental verbs were identified via UCREL Semantic Analysis System (USAS). Findings uncovered significantly high frequencies of mental verbs know, want, and love in English song lyrics through 1960s until 1990s. These three mental verbs possess high inclination to occur alongside personal pronouns I and you, depict social actions, high predilection for simple present tense, and simple sentence structure. These attributes illuminate that song lyrics emulate spoken English, predominantly the informal conversation register.

Keywords: Corpus Linguistics, Corpus Driven, Computational Corpus Linguistics, Mental Verbs, Song Lyrics. 1. Introduction

This study analysed linguistics incidences of mental verbs in English song lyrics. According to Biber, Conrad and Leech (2002), mental verbs are verbs that refer to internal mental states and activities. In the Corpus of Longman Grammar of Spoken and Written English (LGSWE), the 12 most common

mailto:[email protected]





Asian Journal of University Education (AJUE) Volume 17, Number 1, January 2021

223

lexical verbs namely say, get, go, know, think, see, make, come, take, want, give, and mean occurs over 1000 times in the corpus. These verbs function to convey what speakers and writers feel or think through the expressions of cognition (know and think), emotion (love), desire (want), perception (see and taste), and receipt of communication (read and hear) as categorized by Biber, Connrad, Reppen, Byrd, Helt, Clark, Cortes, Csomay & Urzua (2004, p.30). Based on the descriptions, it can be inferred that mental verbs do not involve physical actions. The mental verbs in the current study were semantically tagged via USAS. This study views song lyrics as a source of authentic texts that are written and sung by native speakers, therefore reflect natural behaviours of mental verbs in English language.

Corpus investigations of verbs in song lyrics, including mental verbs, have been repeatedly cited to mirror linguistics phenomena in the English language. Corpus studies of English song lyrics by Logan, Kositsky, and Moreno (2004), Morrison (2012), Falk (2013), Motschenbacher (2016), and Eiter (2017) have included general verbs namely like, know, think, and want in their analyses. Logan, Kositsky, Ramalingam, Krishnan, Suppiah & Maruthai (2020) and Moreno (2004) uncovered that lexical words know, like, and love are in the top 10 most frequent words in their 15589 song lyrics corpus of five music genres namely Country, Reggae, Newage, Rap and Rock. Morrison (2012) analyzed top 20 most popular English karaoke songs in Japan, and found that similar lexical words such as know, love, want, and like were in the top 10 most frequent content vocabulary (p.80). The one syllable verbs in the corpus points to spoken rather than written language, concluding that the song lyrics contain high frequency of words common in spoken English (Morrison, 2012, p.80). On the other hand, Falk (2013) built a larger corpus of 300 lyrics of only Rock genre that spanned over 5 decades from years 1950 to 1999, called the Falk’s Rock Lyrics Corpus (ROLC) which contains lexical words like, love, and know in the top 10 most common words (p.21). Although the absolute frequencies of verbs are not provided, the ranks present which words occur more frequently than the others. Compared to Logan, Kositsky and Moreno (2004), Morrison (2012) and Falk (2013) built smaller corpora that were easier to manage, but the generated data are insufficient for generalisation. For instance, Morrison (2012) stated that vocabulary effective for communication in his corpus cover only 1476 out of 2000 most frequent words from BNC and COCA (pp.80-81), which means that the corpus size is insufficient to represent the aspect of English language communicative vocabulary. Hence, the current study built DCOESL of approximately 1.2 million words, to be able to generate various findings for accurate representativity. Representativity is relative as it varies between corpora and never absolute (Lindquist, 2009; Ab Karim, 2015). By carrying out a comparative analysis of mental verbs in DCOESL with mental verbs in COCA, the researcher could investigate the usefulness of findings emerged from the current study. This includes the extent to which the findings can be generalized to the language. Consequently, making such analysis would certainly make findings more reliable (Lindquist, 2009, p.43).

More recent larger corpora by Motschenbacher (2016) and Eiter (2017) offer more varied insights to linguistics phenomenon such as collocations, keywords or themes, semantic, sentence structures, and tenses. For example, Motschenbacher (2016) highlighted semantic fields in Eurovision lyrics corpus (ESC-ENG), a song lyrics corpus which amounts to approximately 205000-word tokens. Comparative analysis with G-charts revealed that love is the predominant topic in ESC-ENG, particularly in Pop lyrics. This is because the word love occurs 1170 times in ESC-ENG and 642 times in G-charts or 12463 and 5761 times pmw respectively (Motschenbacher, 2016, p.12). Eiter (2017) built the Innsbruck Corpus of English Pop Songs (ICEPS) of roughly 120000 words. The statistical results were submitted for significant tests against COCA and BNC using the Log-likelihood (G2) test. Non-standard feature of English language like the contraction don’t in ICEPS is very significant (G2

COCA=168.1440, G2

BNC=118.7063, df=1) with p ≤ 0.001) (Eiter, 2017, p.27). Another interesting finding is the discourse marker of lexical item like as a verb occur 123 times. For example, in the phrase classic thing that you like (p.35). The researcher managed to draw a representative conclusion that Pop songs are written-to-be-sung language through the uses of non-standard feature and discourse marker. The results and analyses of the studies are not limited to statistical data, but also supported by qualitative data. Statistical data that are supported with statistical tests of significance and concordances attest the reliability and validity of claims and discussions made based on their research findings. In the current study, the quantitative data in the form of frequency distributions are supported with concordance, collocations, and semantic analyses.


224

This study is driven by the massively and freely accessible authentic corpus data that people globally listen and relate to everyday song lyrics. Previous work on song lyrics by Murphey (1992), Kuhn (1999), Kreyer and Mukherjee (2007), Saarinen (2013), Motschenbacher (2016) and Nishina (2017) have largely focused on common word forms and verbs in general found in English song lyrics. Recent studies of mental verbs in corpora that are built upon other registers of spoken and written English have been carried out (Nordlund, 2008, 2010; Sarudin, Faredza, Intan Safinas & Zulkifli, 2019; Verdaguer, 2010). However, a study that is directly focused on mental verbs in English song lyrics is yet to be executed. This study was carried out in view of this research gap and potential linguistic values of song lyrics in mirroring authentic use of English language by the native speakers. 2. Methodology

2.1 Corpus Description: DCOESL

To build the Diachronic Corpus of English Song Lyrics or DCOESL, 5000 song lyrics in the

1960s until 2000s comprising of 25 songs for every year of each genre, were compiled and organized. In total, the song lyrics comprised one million words. DCOESL is intended to be a massive corpus to be able to represent the aspects of English language. McEnery, Xiao and Tono (2006) stated that a specialized corpus has the tendency to be genre specific. DCOESL is a specialized song corpus, consisting of four popular music genres namely Country, Pop, Rock and Rhythm and Blues (R&B).

Table 1. Description of DCOESL

Genre Year No. of Song

Lyrics for Every Year

No. of Word

Country Pop

Rock R&B

1960

-20

09 25

25 25 25

290 278 357 770 303 828 460 545

TOTAL 50 5 000 1 412 601

2.2 Corpus Compilation

Corpus compilation for this study involved selections of music genres, song lists and song lyrics. Four popular music genres namely Country, Pop, Rock and R&B were chosen primarily because popular music is widely listened to by people. Song lists were selected from Top 100 Billboard Charts, an online extension of the well-known Billboard Magazine. This specific platform was carefully chosen because among the many options of accessible songs, chart songs ranked highly in popularity (North, Hargreaves & Hargreaves, 2004). Lastly, song lyrics to build DCOESL were retrieved from the MetroLyrics (www.metrolyrics.com), a freely available online lyrics database of over more than one million song lyrics from 20000 artists. The song lyrics were stored electronically in the form of text (txt.) form metadata. 2.3 Instruments: Computational Corpus Tools

This study adopted five computational corpus analysis tools comprising of: 1) CLAWS Part-of-Speech (POS) Tagger (Rayson, Archer, Piao, & McEnery, 2004), 2) Lancaster Statistics Tools (B�ezina, 2018). 3) LancsBox (B�ezina, Timperley & McEnery, 2018), and 4) UCREL Semantic Analysis System (USAS) English Tagger (Rayson, Piao & Archer, 2004).

First, LancsBox is a fresh generation software package of corpora developed at Lancaster

University. LancsBox was downloaded from its official website (corpora.lans.ac.uk/lancsbox) and installed for free. The current study utilized this toolbox for generating frequency counts, concordance

http://www.metrolyrics.com/


225

lines, collocations and collocational graph (collograph) of mental verbs in DCOESL. Second, CLAWS POS Tagger functions were used to tag words according to their parts of speech. In this study, the C7 tagset was used to tag parts of speech in song lyrics, including verbs, which are tagged as VV0 (base form of lexical verbs), VVD (past tense of lexical verbs), VVZ (-s form of lexical verb), VVG (-ing participle of lexical verb), VVN (past participle of lexical verb). Next, USAS English Tagger served to identify semantic categories of top 20 most frequent lexical verbs in DCOESL. Similar to CLAWS POS Tagger, USAS was developed with its own tagset of 21 major discourse fields. Last, Lancaster Statistics Tools online was accessed to automatically calculate the complex formulae of three statistical tests of significance namely Log-likelihood (G2), Mutual Information (MI), and t-score. As asserted by McEnery, Xiao and Tono (2006), computers help to process and manipulate corpus data rapidly at minimal cost, avoid human biasness to achieve reliability and allow further automatic processing to be performed for various metadata enrichment.

2.4 Triangulation: Reliability and Validity

To guarantee the reliability and validity of statistical findings, five measures were taken: 1) normalization of frequency counts, 2) comparative frequency counts, 3) Log-likelihood (G2), 4) Mutual Information (MI), and 5) T-score

Normalization of raw frequencies in DCOESL to per million words (pmw) as the total

occurrences of verbs in every genre are different. It is more helpful to state how many times a verb occurs, in average, in every one million words. Normalized frequency also helps to accurately compare two different corpora of different sizes. For instance, frequencies of mental verbs in DCOESL were compared to the ones in the COCA, using normalized frequencies as per million words for both corpora.

Comparative frequency counts with general reference corpus COCA were carried out to ensure the representativeness of DCOESL to English language. Through conducting comparative frequency counts analysis, the usefulness of findings from this study was uncovered, allowing for generalisability to the target language. Subsequently, carrying out an analysis as such would result in more reliable findings (Lindquist, 2009, p.43).

Log-likelihood (G2) test was carried out for comparison of frequency counts between DCOESL and COCA to know if the statistical differences were due to real world differences or just by chance. G2 is broadly exploited as a measure of strength of associations (Moore, 2004). The G2 for this study was calculated based on normalized frequencies. Number of occurrences of mental verbs in DCOESL were compared to the ones in COCA to recognize on a quantitative basis the most significant contrasts between the two lists (Allan, 2015).

To test collocational strength of mental verbs collocations, MI test was carried out. An MI score of 3 or higher is to be taken as an evidence that two items are collocates (Hunston, 2002, p.71). Thus, if MI = < 3, H0 is accepted, and if MI = > 3, H0 is rejected. Meanwhile, if MI = > 3, Ha is accepted, and if MI = < 3, Ha is rejected. Meanwhile, t-score is applied to validate MI scores by giving clearer understanding to which words have a strong attraction to the mental verbs and which do not occur frequently in DCOESL are not given a high significance. Collocations with t-score of 2 or higher should be considered as important (Hunston, 2002, p.72).

To be certain that the collocations in the current study are the results of more than vagaries, another collocation measurement was taken into measure; the T – score. T – score is utilized to analyze and validate MI scores by giving clearer insight to which words have a strong attraction to the lexical verbs and which do not occur frequently in DCOESL are not given a high significance. Collocations with T – score of 2 or higher should be considered as significant (Hunston, 2002, p.72). 3. Results

The investigation comprises of four types of findings: frequency list, collocations, concordances and comparative analysis. First, the frequency list contained 20 most common general verbs in DCOESL.


226

Secondly, the general verbs were semantically tagged to identify mental verbs and to select the top three mental verbs. Next, the selected mental verbs were submitted for comparative analysis with Corpus of Contemporary American English (COCA). Last, the collocations of the mental verbs were identified, and tests for collocational strength were performed. These data was utilized to form discussion and to draw conclusion for the current study. 3.1 General Verbs

Table 2 presents the generated top 20 general verbs in DCOESL and their respective G2 test

values against COCA. Table 2. G2 Test Results of Top 20 General Verbs in DCOESL against COCA

DCOESL vs. COCA

Rank Word NF G2 Sig. Level

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

know want love come get say let go take see tell make think like give keep look hold shake call

3562 2372 2206 1928 1806 1449 1428 1347 1085 1063 993 973 853 811 753 733 519 411 408 383

1447.02 1292.53 2239.10 1496.65 998.36 594.54 981.47 690.08 610.10 277.06 751.86 495.65 0.01 459.60 524.37 537.65 83.01 319.54 504.28 150.61

<0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.1 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

Note. Values of significance as given by McEnery, Xiao, and Tono (2006, p.55): Null hypothesis, H0 = There exists no significant association between the occurrences of general verbs in DCOESL with those in COCA. Alternative hypothesis, Ha = There exist a significant association between the occurrences of top 20 general verbs in DCOESL with those in COCA (G2>6.63 at p < 0.01 or 1% level, G2> 10.83 at p < 0.001, G2> 15.13 at p < 0.0001).


227

3.2 Mental Verbs

Table 3. Semantic Categories of Know, Want, and Love in DCOESL

Word POS Tag

USAS Semantic Tag

Category

know VV0 X2.2+ Mental actions and processes: Knowledge/Perception/Retrospection

S3.2 Relationship: Intimate/ Sexual

B1% Anatomy and physiology

want VV0 X7+ Depicts desire or aspiration.

S6+ Obligation and necessity.

A7+ Definite (+ modals). Abstract term of modality to show possibility, necessity, certainty, etc.

S3.2 Terms relating to relationships that are intimate and/or sexual.

A9- General or abstract terms relating to allocating/ relinquishing/ acquiring/ receiving, etc.

love VV0 E2 : Liking

E2+ : Fondness/affection/partiality/attachment

Note. Antonymity of conceptual classifications is indicated by +/- markers on tags, whereby “+” refers to positive meanings and “–“ refers to negative meanings.

Mental verb know in DCOESL is ranked first among the top 20 general verbs in base forms

with 3562 instances pmw (see Table 2). This verb also demonstrates high G2 values against COCA (G2

COCA = 1447.02, G2 COCASPOKEN = 2061.55, and G2

COCAWRITTEN = 3132.11 at p < 0.0001) denoting significant relationship between know in DCOESL and in COCA. Semantically tagged via USAS English Tagger, lexical verb know in DCOESL is mainly coded as 'mental activity which is knowledge (X2.2). This semantic meaning of know is similar to the descriptions of the lexical verb given by general reference grammars such as Biber et al. (1999), Hornby, Turnbull, Lea, Parkinson, Phillips, Francis, Webb, Bull and Ashby (2010), and Stevenson (2010) whereby summarized, these linguists described the verb know as an individual’s mental activity of having information about something (see Table 3). Other semantic categories of the verb are relationship (S3.2) and human body (B1). Many songs are about romantic relationships between romantically involved individuals (Paxson, 2013, p.35), family relationship (Homan, 2006), and contain sexual related contents (Galician & Merskin, 2007, p.125). Thus, it makes sense that the verb know in DCOESL somehow carries meanings of relationships and human body.

Mental verb want in DCOESL is ranked second with 2372 instances pmw (see Table 2). This mental verb possesses high G2 values (G2

COCA = 1292.53, G2 COCASPOKEN = 2045.24, and G2

COCAWRITTEN = 1872.61at p < 0.0001). Lexical verb want in DCOESL is generally used to express positive desire or aspiration (X7+). Other semantic uses of want in DCOESL are to demonstrate social actions, states and processes (S6+ and S3.2), act as an abstract term of modality to show necessity (A7+), and act as a


228

general or an abstract term to show possession (A9-). Social function of want relates to intimate or sexual, or both in relationships (S3.2). Want in DCOESL is a social expression to show obligation, necessity, and/ or desire, particularly those related to relationships. This resounds general reference grammars of English like Biber et al. (1999, p.378) and Hinkel (2004, p.191) who asserted that want is a common emotive or mental verb which expresses meanings such as obligation and necessity. Pinna and Tortorici (2010, p.404) also stated that lexical verb want is categorized as one of the common verbs that relates to the meaning of desire. Song lyrics are a medium of expression to express emotions such as anger, calmness, happiness and sadness (Jamdar, Abraham, Khanna & Dubey, 2015, p.42), as well as desire as reported by Bridle (2018, p.26) in his corpus analysis of R&B lyrics. Note that want is found to be predominant in the R&B genre of DCOESL (966 instances pmw). Considering this, it is not random that song lyrics in DCOESL also express the same emotion as identified through semantic tagging of the emotive or mental verb want.

Mental verb love is ranked third with 2206 instances pmw (see Table 2), with significant G2

values (G2COCA = 2239.10, G2

COCASPOKEN = 2629.14, and G2 COCAWRITTEN = 2496.97 at p < 0.0001). Based

on the generated results, it is clear that mental verb love in DCOESL is mainly used to express liking, which relates to the feeling or emotion of affection.

Fig. 1 Occurrences of Know, Want, and Love across Time in DCOESL

The overall frequencies of lexical verb know increase approximately 81.32 per cent over time from 1960s to 2000s in DCOESL. The occurrences of know increase by about 35.04 per cent in 1970s but decrease by about 25.12 per cent in 1980s, and then slightly increase by about 5.4 per cent in 1990s. From 1990s, instances of know increase by about 49.22 per cent. The chi-square (X2) test confirms that the differences in frequencies of know in DCOESL across genre and time are highly significant (X2 = 508.054, with p-value=0.0001, the result is significant at p<0.01). The Log-likelihood (G2) test for know in DCOESL against reference corpus COCA and their subcorpora also confirms that the occurrences of know among the corpora are not due to randomness. A total of approximately 36.36 per cent of know occurs in RnB of DCOESL, followed by Pop, Rock, and Country with 27.65 per cent, 20.58 per cent, and 15.41 per cent respectively.

530

816

611 644

961

296 316

447

668 648

529447 423 393 415

0

200

400

600

800

1000

1200

1960-1969 1970-1979 1980-1989 1990-1999 2000-2009

FR

EQ

UE

NC

Y P

ER

MIL

LIO

N W

OR

DS

(P

MW

)

YEARS

Know Want Love


229

For mental verb want, there is a noticeable difference in the frequencies of lexical verb want over time for the four popular music genres in DCOESL. The occurrences of want gradually increased from 1960s by approximately 6.76 per cent in 1970s, 41.46 per cent in 1980s, 49.44 per cent in 1990s, and slightly decrease by about 2.99 per cent in 2000s. X2 of 142.67 shows highly significant result at p < 0.001. The G2 scores against the reference corpora and their subcorpora also gives results as highly significant particularly for the spoken register of COCA (G2

COCA=1292.53 at p<0.0001, G2COCASPOKEN =

2045.24 at p<0.0001, and G2COCAWRITTEN=1872.61 at p<0.0001). The occurrences of lexical verb want

is very dominant in RnB with 40.73 per cent from the total occurrences of the verb, followed by Pop (29.43 per cent), Rock (15.89 per cent), and Country (13.95 per cent). The frequencies of love show that there is a discernible difference in the overall use of the lexical verb love over time in DCOESL. The occurrences of love gradually decrease at an average 9.32 percent from the 1960s until 1980s, and then slightly increase by 5.6 percent in the 2000s. The chi square (X2) test verifies that the differences over time are highly significant (X2 = 118.8215 at p < 0.00001). Log-likelihood (G2) test of DCOESL against the COCA and its sub corpora also reveals that the occurrences of love in the corpora are highly significant (G2

COCA = 2239.10 at p < 0.0001). To investigate the linguistics structures of mental verbs know, want, and love in DCOESL, the

adjacent collocations immediately to the left (window span: -1) and right (window span: +1) of the verb were identified via Lanscbox and the submitted for statistical tests of significance to calculate for Mutual Information (MI) scores and T-scores. This is to access the importance of the collocations and show a clearer picture of the relationship between words than that given by a simple collocation list itself. 4. Discussion

Hornby et al. (2010) defined the lexical verb know as “to have information in your mind as a result of experience or because you have learned or been told it” and also refers to “realize, feel certain, be familiar, recognize, distinguish, skill or language, and experience” (p.826). In other words, to know something is to have information in one’s mind about that matter in context as in sample lyrics (1) and (2). For want, Waite (2013) described the lexical verb as “to feel a need or desire to have or do” (p.1044), and Hornby et al. (2010) interpreted it as “to have a desire or wish for something” (p.1672). Simply put, to want something means to have desire or longing to possess that specified thing one has in mind as in sample lyrics (3) and (4). Lexical verb love is interpreted as to “feel a deep romantic or sexual attachment to” and “like or enjoy very much” (Stevenson & Waite, 2011, p.845). Otherwise stated, to love is by having a relationship or connection, which brings meaning to the people in that liaison as shown in sample lyrics (5), (6), and extreme feeling of liking towards someone or something as in (7).

Fig. 2 Concordance Lines of Selected Lyrics for Know, Want, and Love in DCOESL

Verbs know, want, and love are classified as mental or emotive activity verbs apart from think

and see (Hinkel, 2004, p.189; Pearce, 2012, p.113). Song lyrics play a massive role in music as a vessel to express emotions, evaluations, and values to the mass audiences (Kennedy & Gadpaille, 2014, p.50). Hence, there is no surprise that in DCOESL, lexical verbs know, want, and love are ranked first, second and third respectively. 4.1 Language Uses of Mental Verbs


230

Instead of merely focusing on the three mental verbs alone, their adjacent collocations assist in giving insights on the language uses of know, want, and love. For instance, the to-infinitive in DCOESL occurs immediately to the right of verb want, which gives characteristic to want from typical lexical and active verb to semi modal. Real linguistic structure could be identified for the prominent verbs, instead of focusing solely on their ‘statistical phenomenon’ (Lindquist, 2009, p.78).

Fig. 3 Collograph of Know, Want, and Love

Table 4. Top Three Adjacent Collocations of Know, Want, and Love

Mental Verbs

Rank Collocate at -1

NF MI T Collocate at +1

NF MI T

Know 1. 2. 3.

I you don’t

1779 1654 740

3.577 3.635 5.020

38.643 37.395 26.364

i you that

812 675 594

2.445 2.342 5.550

23.263 20.855 23.852

Want 1. 2. 3.

I you don’t

1316 760 580

3.728 3.099 5.255

33.54 24.352 23.452

to you me

1080 410 220

4.991 2.209 2.451

31.830 15.869 12.119

Love 1. 2. 3.

I you really

1117 367 50

3.597 2.154 7.339

30.659 14.852 7.027

you me the

812 355 156

3.3 3.246 4.217

25.602 16.855 11.818

Note. The raw frequencies have been normalized to as per million words (pmw) for uniform representation. The formula applied is (NRF x 106) / T = NNF whereby, NR = value of the raw frequency of base form of lexical verb, T = total tokens of the corpus (1412601) and NNF = value of the normalized frequency. The statistical test values are significant at MI ≥ 3 and T ≥ 2 (Hunston, 2002, pp.71-72). To further investigate the phenomena of pronouns I and you, with mental verbs know, want, and love in DCOESL, top three collocations in DCOESL and COCA were searched for and the findings are as shown in Table 5.

Table 5. Top Three Adjacent Collocates to the Left (Window Span -1) and Right (Window Span +1) of Personal Pronouns I and You in DCOESL

Threshold=150.0


231

DCOESL and the reference corpus emphasize on patterns and + I (conjunction + pronoun), that + I (conjunction + pronoun), when + I (conjunction + pronoun), if + you (conjunction + pronoun), love + you (verb + pronoun), that + you (conjunction + pronoun), I/ you + know (pronoun + verb), I/you + want (pronoun + verb), and I/you + love (pronoun + verb). As can be seen in Table 5, some of these patterns are found in the reference corpus. The pronoun + verb pattern is very significant in DCOESL and elucidated in the succeeding paragraphs.

Murphey (1992) stated that songs reflect what some listeners want to say by literally placing the words into their mouths as they sing along. This is because songs are relatable to listeners and the lyrics carry stories from the artists to be shared or sung with listeners, permitting them to engage in the songs (Griffee, 1992, p.4). Below are five examples of lyrics in DCOESL containing personal pronouns I and you.

Rank DCOESL COCA COCA

SPOKEN

COCA

WRITTEN W

ind

ow

Sp

an

-1

1 and when wish When

2 that because whenever If

3 when guess interpreter Think

1 if if if If

2 love when thank When

3 that thank when What

Win

dow

Sp

an

+1

1 know think think Think

2 want know mean Know

3 love want want Want

1 know know know Know

2 want think want Want

3 love want need Think

Note. Auxiliary and modal verbs have been excluded from the collocation search function for all the corpora, as the items are not of main interest in this research.


232

Fig. 4 Concordance Lines of Selected Lyrics for I and You in DCOESL

According to Murphey (1992), the fact that I in song lyrics is anonymous, it becomes easier for listeners to “appropriate the words” (p.171). In other words, listeners seize to be the subject I, just as if he or she is communicating with the object you, in a context such as conveying perception (know) in (1), (2), and (3), and emotional connection (love) between the subject and object in (4). In other words, personal pronouns I and you are used by the subject and object in an interaction regarding their views and feelings (Ha, Ariffin & Ma’rof, 2018). This finding is almost similar to the linguistics construction of Eurovision Song Contest (ESC) found by Motschenbacher (2016) which involved three components namely the desiring subject (I), the desired object (you), and the relationship between the two (love) (p.203). The listeners identify themselves with the lyrics, wherein the personal pronouns serve as a basic form of personal reference (Sanchez-Stockhammer & Schubert, 2016; Baharum, Ariffin & Abd Wahab, 2019).

Table 6. Samples Sentences Involving Pronouns I and You with Lexical Verb Know

No. Sample

Sentence/Phrase

DCOESL COCA

Spoken Written

RF NF NF G2 NF G2

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16.

I know You know I don’t know I know you I know I I know that You don’t know I know what I know the I know how I don’t even know I know why I know where I don’t know you I don’t know that You don’t know that

2513 2337 542 439 375 275 225 149 69 64 35 32 19 3 2 2

1779 1654 384 311 265 195 159 105 49 45 25 23 13 2 1 1

80 584 3 9 3 11 0.3 3 2 2 0.1 0.3 0.4 0.01 0.2 0.01

1917.29 533.10 501.36 361.59 338.61 199.72 216.47 122.30 53.83 48.61 33.49 29.09 14.98 2.66 0.58 1.29

113 134 0.2 9 7 10 0.04 7 4 4 0 1 1 0 0 0

1766.89 1526.61 529.19 361.59 312.02 204.28 219.73 102.90 45.11 40.22 34.66 24.96 12.20 2.77 1.39 1.39

Average Man G2 Scores 273.436 324.119 Previous studies on song lyrics corpus data by Kreyer and Mukherjee (2007) and Falk (2013)

focusing on Pop and Rock genres respectively, discovered extremely high frequencies of personal pronouns I and you in song lyrics. Kreyer and Mukherjee (2007) explored their Giessen-Bonn Corpus of Popular Music (GBoP) and uncovered that personal pronouns I and you have relative frequencies of 3.78 per cent and 3.87 per cent respectively (p.44). Falk (2013) built the Rock Lyrics Corpus (ROLC) to compare the corpus with GBoP and found that ROLC also contains extremely high frequencies of personal pronouns I and you covering 7.31 per cent and 5.99 per cent correspondingly of the overall words in the corpus (p.13). The corpus studies on Blues genre by Kuhn (1998) uncovered that Blues lyrics resembles real life requests (p.527). Bridle (2018) in the Male Blues Lyrics is focused to investigate language use in the pre-World War Two and post-World War Two corpora, and stated that pronouns like I and you (p.28) in the corpora are related to category of Relationship: Intimacy and Sex (p.27). For DCOESL, the personal pronouns I and you occur about 4.19 per cent and 3.74 per cent respectively overall in the corpus. This study unveils that the two pronouns are extremely high in the RnB genre with 34.27 per cent occurrences of the total I and you in the corpus. Overall, I and you in DCOESL have the highest resemblance to spoken register when compared to general reference corpus COCA.


233

Table 7. Distribution of Know, Want, and Love Variants in DCOESL and G2 Test Results against COCA

DCOESL vs. COCA Form POS Tag Genre NF G2 Sig. Level know VV0 Country 549

1447.02

<0.0001

Pop 985 RnB 1295 Rock 733 3562 knew VVD Country 114

678.33

<0.0001

Pop 116 RnB 166 Rock 94 490 knows VVZ Country 94

458.12

<0.0001

Pop 99 RnB 83 Rock 55 331 knowing VVG Country 25

96.81

<0.0001

Pop 27 RnB 24 Rock 8 84 known VVN Country 37

114.26

<0.0001

Pop 36 RnB 25 Rock 43 141 want VV0 Country 331 1292.53 <0.0001 Pop 698 RnB 966 Rock 377 2372 wanted VVD Country 42 17.21 <0.001 Pop 48 RnB 41 Rock 56 187 wants VVZ Country 38 19.04 <0.0001 Pop 58 RnB 40 Rock 88 224 wanting VVG Country 10 3.39 <0.05 Pop 16 RnB 9 Rock 6 41 wanted VVN Country 17 13.02 <0.001 Pop 30 RnB 21 Rock 11 79 love VV0 Country 418 2239.10 < 0.0001


234

Pop 684 RnB 848 Rock 256 2206 loved VVD Country 59 30.20 < 0.0001 Pop 44 RnB 35 Rock 30 168 loves VVZ Country 88 250.11 < 0.0001 Pop 107 RnB 59 Rock 35 289 loving VVG Country 76 301.35 < 0.0001 Pop 55 RnB 102 Rock 24 257 loved VVN Country 41 101.47 < 0.0001 Pop 30 RnB 32 Rock 27 130 Note. Values of significance as given by McEnery, Xiao, & Tono (2006, p.55): Null hypothesis, H0=There exists no significant association between know, want, and love variants in DCOESL with those in the reference corpora. Alternative hypothesis, Ha=There exists a significant association between the occurrences know, want, and love variants in DCOESL with those in the reference corpora (G2> 10.83 at p < 0.001, G2> 15.13 at p < 0.0001).

Based on the quantitative and qualitative findings that are discovered from DCOESL, it is

assumed that it is the nature of English song lyrics to reflect immediate communication similar to that of face-to-face conversation. This assumption is grounded upon six characteristics of English song lyrics that emerge from DCOESL; 1) high density of cognition and mental verbs, 2) high density of pronouns, 3) use of contraction, 5) simple sentence structure, and 6) high density of simple present tense.

Daniel J. Levitin, an American Canadian cognitive psychologist, a neuroscientist, writer, musician, and record producer, stated that knowledge and emotion are intertwined. He disagreed with the stereotypical perspective that science consists solely of facts and measurements that are isolated to emotion. He described that through emotional judgment, artists selected what they thought were vital and transferred their observations in life into an intelligible complete work of art called music (Levitin, 2019). Therefore, making music is an emotional journey for artists. Music can be referred to as a single coordinated pattern of neurochemical impulses in the brain (Rowell, 1984, p.1), and is generally defined by Hornby et al. (2010) as “sounds that are arranged in a way that is pleasant or exciting to listen to” (p.972). However, music is incomplete without song writing (Hauser, Tomal & Rajan, 2017). This team of musical composer, guitar rocker, and lyric opera singer describe song lyrics as:

The lyrics reflect our feelings and emotions beyond the music, tell us stories, and give us new way of thinking about everyday experiences. The lyrics translate the aural experience of a song into something tangible connecting the movement of the melody and enabling the music to become personal and memorable. … The relationship between music and lyrics is at the very heart of good songwriting. (Hauser, Tomal, & Rajan, 2017, p.85)


235

Communication is defined as the “activity or process of expressing ideas and feelings or giving people information” (Hornby et al., 2010, p.290). To communicate is to share ideas or feelings. People delve into songs for many reasons including a compelling need for self-expression and a desire to communicate something to others. Kennedy and Gadpaille (2014) stated that song lyrics is a medium to express emotions, evaluations, and values to the audience (p.50). Through song lyrics, artists have a medium to carry and communicate their stories to the audience (Griffee, 1992). Therefore, English song lyrics utilize high frequencies of mental verbs namely know, want, and love as demonstrated in DCOESL. This is naturally appropriate because when writing song lyrics, it is very important to remind oneself that song writing is synonymous with emotions, and successful songs tap into the emotions that everyone feels (Hirschhorn, 2001, pp.29-31). Otherwise stated, song lyrics must be emotional.

Findings also show that DCOESL contains extremely significant frequencies of pronouns I and you. In fact, I and you is also at the top two most common word forms in DCOESL. The use of personal pronoun I helps listeners to appropriate the words (Murphey, 1992, p.171), which places the listeners as the subject (I) who is communicating with the object (you). This leads to the linguistics construction of Subject + Verb + Object, which is the primary sentence structure in DCOESL. This finding is similar to the finding made by Motschenbacher (2006) in his corpus of ESC-ENG.

Other than that, the use of contraction don’t is common in DCOESL. Contraction don’t is very prominent whereby it can be found as a collocate directly to the left of know and want (see Table 4). Contractions are usually associated with speech and have so far normally been avoided in formal writing (Cheng, 2002, p.13; Nasroniazam, 2016). DCOESL contains high frequencies of contraction don’t because lyrics are a medium of free expression and therefore does not follow the formal conventions of written English. Baker (2009) stated that the use of contractions is to lessen the processing burden for the listener. For instance, a song can last up to three minutes. An artist has to pack the lyrics to make sure that the song lyrics convey stories and emotions within that short period of time to listeners. Contractions provide immediacy to content delivery like how it is done naturally in informal speech such as conversation. This is because contractions keep a sentence short and makes content delivery to consume less time. Overall, English song lyrics portrays interactive features through high preference for mental verbs, personal pronouns, simple sentence structure, and present tense. These features are very similar to findings by corpus linguistics researchers such as Biber et al. (1999), Motschenbacher (2016), Kreyer & Mukherjee (2007), Nishina (2017), Roland, Dick & Elman (2007), Eiter (2017) and Lovett (2020) to name a few. 4.2 Usage-Based Model of Mental Verbs in DCOESL

The extremely high instances of personal pronouns alongside mental verbs in DCOESL demonstrates great emphasizes on cognition and emotional activities between singers or songwriters, and the person the person they direct the song lyrics to. For the purpose of general representation, singers and songwriters in this study are called as the artists. Motschenbacher (2016) described this type of linguistics structure (SVO) as the relationship between the desiring subject and the desired object (p.203). In DCOESL, an artist (represented by I) direct his or her attention to the person that causes the mental state. The artist addresses the person involves in the context of the song. Thus, the artist becomes the addressor, and automatically the person being mentioned or targeted (anonymously or known) in that particular context or relationship is called as an addressee. Figure 5 outlines the lexical system for lexical verbs know, want, and love in DCOESL. The structure demonstrated is the main basic SVO structure found in DCOESL. This structure can be inflected for example by adding clausal item that, and reduction such as contracted words don’t and won’t. The terms addressor and addressee in DCOESL agrees with the representation made by Pustejovsky (1993) on the two-way causal relation of mental state whereby the terms experiencer and stimulus are used instead.


236

Fig. 5 Usage-based Model of Mental Verbs (Know, Want, and Love) in DCOESL

Grounded upon the usage-based conception developed by Langacker (1991), Mukherjee (2005) specified four central principles of a usage-based model namely (1) a real-data model of natural language data, (2) a frequency-oriented model based on quantitative analysis of representative corpus data, (3) a lexicogrammatical model that overcome the traditionally established boundary between lexis and syntax, and (4) a unified core-periphery model that reflect language use by actual speakers in natural contexts (pp.221-224). The usage-based model of mental verbs in Figure 3 is a lexical framework centered on (1) native speakers’ song lyrics, (2) lexical items that were submitted to statistical tests of significant against large general reference corpus, (3) combine a set of lexical items with their grammatical structure, and (4) pointed that song lyrics possess similar characteristics of spoken features, making them a written-spoken-like genre.

Pustejovsky (1993) explained that two processes are involved in obtaining a mental state namely first, the experiencer must direct his or her attention to the stimulus and second, the stimulus causes the experiencer to come into a certain mental state (p.64). Hence the two-way causal relation. In DCOESL, the spoken-like interaction between artists and the persona (or thing) in their song lyrics are defined in terms of the addressor- addressee relationship or context.

5. Conclusion

This study identified most frequent verbs, most frequent mental verbs, and analyzed the uses of mental verbs in English song lyrics of four music genres through 1960s until 2000s. Overall, five features of English song lyrics are identified based on this diachronic analysis of DCOESL; 1) high density of cognition and mental verbs, 2) high density of pronouns, 3) use of contraction, 4) simple sentence structure, and 5) high density of simple present tense. A usage-based model of mental verbs know, want, and love was developed to reflect the findings from this study. These interesting findings ultimately lead to the conclusion that song lyrics are conversational, or a conversation-like register.

6. Acknowledgement

The authors wish to thank Universiti Pendidikan Sultan Idris (UPSI) for supporting this

research and the Research Management and Innovation Centre (RMIC) UPSI for funding the journal publication.


237

7. References

Ab Karim, R. (2015). Huraian Kata Adverba Bahasa Melayu Berdasarkan Data Korpus Berkomputer. PENDETA Journal of Malay Language, Education and Literature, 5, 91–112.

Allan, Q. G. (1999). Enhancing the language awareness of Hong Kong teachers through corpus data: The TeleNex experience. Journal of Technology and Teacher Education, 7(1), 57-74.

Anida Sarudin, Husna Faredza, M. R., Intan Safinas, M. A., & Zulkifli, O. (2019). Using the Cognitive Research Trust Scale to Assess the Implementation of the Elements of Higher-Order Thinking Skills in Malay Language Teaching and Learning. International Journal of Recent Technology and Engineering, 8(2S2), 392–398. https://doi.org/10.35940/ijrte.B1064.0782S219

Baker, P. (2009). Contemporary Corpus Linguistics. London, United Kingdom: Continuum International Publishing Group.

Biber, B., Connrad, S. M., Reppen, R., Byrd, P., Helt, M., Clark, V., Cortes, V., Csomay, E., & Urzua, A. (2004). Representing language use in the university: Analysis of the TOEFL 2000 spoken and written academic language corpus. Princeton, New Jersey: Educational Testing Service.

Biber, D., Conrad, S., & Leech, G. (2002). Longman student grammar of spoken and written English. Harlow, United Kingdom: Pearson Education Limited.

B�ezina, V. (2018). Statistics in corpus linguistics: A practical guide. Cambridge, United Kingdom: Cambridge University Press.

B�ezina, V., Timperley, M., & McEnery, T. (2018). #LancsBox v.4.x [software]. Available at :http://corpora.lancs.ac.uk/lancsbox.

Bridle, M. (2018). Male Blues lyrics 1920 to 1965: A corpus based analysis. Language and Literature, 27 (1), 21-37.

Cheng, W. (2012). Exploring corpus linguistics: Language in action. Oxfordshire, United Kingdom: Routledge.

Eiter, A. (2017). Haters gonna hate: A corpus linguistic analysis of the use of non-standard English in Pop songs (Bachelor Thesis). University of Innsbruck. Retrieved from https://www.researchgate.net/publication/318094466

Falk, J. (2013). A diachronic corpus-based analysis of linguistic features in rock lyric. Galician, & Merskin, .(2007). Critical thinking about sex, love, and romance in the mass media:

Media Literacy Application. New Jersey: Lawrence Erlbaum Associates, Inc.. Griffee, D. T. (1992). Songs in action. Hertfordshire: Phoenix ELT. Ha Ming Hui, Aryantie Ariffin, & Aini Marina Ma’rof. (2018). An Analysis of Lexical Negative

Transfer in English Writing of Malay Students. AJELP: The Asian Journal of English Language and Pedagogy, 6, 56–64. https://doi.org/10.37134/ajelp.vol6.5.2018

Hauser, C. V., Tomal, D. R., & Rajan, R. S. (2017). Songwriting: Strategies for musical self-expression and creativity. Maryland, United States: Rowman & Littlefield.

Hinkel, E. (2004). Teaching academic ESL writing: Practical techniques in vocabulary and grammar. New Jersey, States: Lawrence Earlbaum Associates.

Hirschhorn, J. (2001). The complete idiot’s guide to songwriting. Indianapolis: Pearson. Homan, S. (2006). Access all eras: Tribute bands and global pop culture. England: McGraw-Hill

Education. Hornby, A. S., Turnbull, J., Lea, D., Parkinson, D., Phillips, P., Francis, B., Webb, S., Bull, V, &

Ashby, M. (2010). Oxford advanced learner's dictionary international student's edition (8th ed.). New York: Oxford University Press.

Hunston, S. (2002). Corpora in Applied Linguistics. Cambridge: Cambridge University Press. Jamdar, A., Abraham, J., Khanna, K., & Dubey, R. (2015). Emotion analysis of songs based on

lyrical and audio features. International Journal of Artificial Intelligence & Applications (IJAIA), 6(3), 35-50.

Kennedy, V. & Gadpaille, M. (2014). Words and music. Newcastle, United Kingdom: Cambridge Scholars Publishing.

Kreyer, R. & Mukherjee, J. (2007). The style of Pop song lyrics: A corpus linguistics pilot study. Journal of English Philology, 125(1), 31-55.


238

Kuhn, E. D. (1999). I just want to make love to you' - Seductive strategies in Blues lyrics. Journal of Pragmatics, 31, 525-534.

Langacker, R. W. (1991). Cognitive linguistics research. Berlin, Germany: Mouton de Gruyter. Levitin, D. (2019). The world in six songs: How the musical brain created the human nature. United

Kingdom: Penguin. Lindquist, H. (2009). Corpus linguistics and the description of English. Edinburgh: Edinburgh

University Press. Logan, B., Pedro, J. M., & Kositsky, A. (2004). Semantic analysis of song lyrics., presented at IEEE

International Conference on Multimedia and Expo, Taipei, Taiwan, 2004. Cambridge, IEEE. Lovett, T. W. (2020). Gendered teaching identities and Chinese undergraduate students’ perceptions

of teacher efficacy in the english language classroom. Asian Journal of University Education, 16(2), 196–204. https://doi.org/10.24191/AJUE.V16I2.7854

McEnery, T., Xiao, R., & Tono, Y. (2006). Corpus-based language studies: An advanced resource book. Oxon: Routledge.

Moore, R. C. (2004). On Log-Likelihood-ratios and the significance of rare event. Paper presented at the EMNLP Proceeding.

Morrison, B. (2012). An analysis of lexis in English-medium songs popular in Japan. Studies in Linguistics and Language Teaching, 23, 71-84.

Motschenbacher, H. (2016). Language, Normativity, and Europeanisation: Discursive Evidence from the Eurovision Song Contest. Coventry, United Kingdom: Palgrave Macmillan.

Murphey, T. (1992). The discourse of Pop songs. TESOL Quarterly, 26 (4), 770-774. Nasroniazam, A. B. S. (2016). Peluasan Makna Kata Nafi Bukan Dan Tidak : Analisis Korpus

Bahasa. PENDETA Journal of Malay Language, Education and Literature, 7(1), 73–84. Retrieved from https://ejournal.upsi.edu.my/article/2017AR001402

Nishina, Y. (2017). A study of Pop songs based on the Billboard corpus. International Journal of Language and Linguistics, 4(2), 125-134.

Norzie Diana Baharum, Kamisah Ariffin, & Rahimah Abd Wahab. (2019). Code mixing by a Non-ESL content instructor: The Language Choice and Syntactic Features. AJELP: The Asian Journal of English Language and Pedagogy, 7(1), 25–37.

Nordlund, M. (2008). From physical to mental acquisition: A corpus-based study of verbs. (Published doctoral dissertation, Luleå University of Technology). Retrieved from https://www.diva-portal.org

Nordlund, M. (2010). Meaning extensions of grasp: A corpus-based study. Journal in English Lexicology, 4, 105-126. doi: 10.4000/lexis.603

North, A. C., Hargreaves, D. J., Hargreaves, J. J. (2004). The uses of music in everyday life. Music Perception, 22, 63–99.

Paxson, P. (2003). Mass communication and media studies: An introduction. New York, USA: The Continuum International Publishing Group Inc..

Pearce, M. (2012). The routledge dictionary of English language studies. Routledge. Pinna, A. & Tortorici, I. (2010). A corpus-driven study of the frame I want to + verb in U.S.

presedential speeches (1993-2010). AnnalSS, 7. Pustejovsky, J. (1993). Semantics and the lexicon. United States: Springer Science and Business

Media. Ramalingam, S., Krishnan, I. A., Suppiah, P. C., & Maruthai, E. (2020). Word Order in Job

Interviews: The Malaysian Perspective. Asian Journal of University Education, 16(3), 148. https://doi.org/10.24191/ajue.v16i3.11080

Rayson, Archer, Piao, & McEnery (2004). CLAWS part-of-speech tagger for English. Retrieved November 23, 2017 from http://ucrel.lancs.ac.uk/claws/

Rayson, P., Piao, S., & Archer, D. E. (2004). The UCREL semantic analysis system. Retrieved November 23, 2017 from http://ucrel.lancs.ac.uk/usas/

Rowell, L. (1984). Thinking about music: An introduction to the philosophy of music. United States: University of Massachusetts Press.

Saarinen, E. (2013). From war pigs to unsung heroes: The criticism and justification of war in metal lyrics. (Master Thesis) University of Turku. Retrieved from http://www.semanticscholar.org

https://doi.org/10.24191/AJUE.V16I2.7854

https://doi.org/10.24191/ajue.v16i3.11080

http://ucrel.lancs.ac.uk/usas/


239

Sanchez-Stockhammer, C. & Schubert, C. (2016). Variational Text Linguistics: Revisiting Register in English. Berlin, Germany: De Gruyter Mouton.

Stevenson, A. & Waite, M. (2011). Concise Oxford English dictionary. New York: Oxford University Press.

Stevenson, A. (2010). Oxford dictionary of English. Oxford: Oxford University Press. Verdaguer, I. (2010). A corpus-based contrastive study of mental verbs. Journal of Semantic

Scholar, 663-686. Retrieved from https://semanticscholar.org Waite, M. (2013). Pocket Oxford English dictionary (11th ed.). Oxford, United Kingdom: Oxford

University Press.

conversational mental verbs in english song lyrics: a

Documents